Genetic and environmental determinants involved in the onset and progression of multiple sclerosis

Multiple sclerosis (MS) is one of the leading causes of neurological disability in young adults. MS occurs in people who have an underlying genetic susceptibility and are exposed to viral and environmental risk factors. The heterogeneity in the clinical presentation of MS has posed a significant challenge for identifying risk factors associated with MS outcomes. The overall aim of this thesis was to examine genetic, lifestyle, and environmental risk factors associated with the onset and progression of MS. Particularly factors associated with relapse risk and risk of worsening of disability, and how these associations can be modified by disease modifying therapies (DMTs).
The Australian Multicentre Case-Control Study of Environment and Immune Functions (the Ausimmune Study), the Ausimmune Longitudinal Study (the AusLong Study), the Tasmanian MS longitudinal Study (the TasMSL Study), and Tasmanian MS Genes and Prevalence Study (the TasMSPG Study), are the data platforms used for the studies presented in this thesis. The Ausimmune Study is a population-based Australian multicentre case-control study designed to capture all incident first demyelination events (FDEs) cases in four regions: Queensland (QLD: latitude 27o south), New South Wales (NSW: 33o south), Victoria (VIC: 37o south), and Tasmania (TAS: 43o south). The Ausimmune Study seeks to investigate the role of environmental, topographic, and lifestyle factors in the development of FDEs, a frequent precursor to MS. These factors included past and recent sun exposure (and vitamin D levels), Epstein-Barr viral (EBV) infections, chemical exposures, dietary exposures, and genetic factors. Between 2003 and 2006, 279 case participants from the Ausimmune Study and another 407 participants from the general population were recruited into the AusLong, TasMSL, and/or TasMSPG studies, and followed up to 15 years post FDE.
The first study, a semi-systematic review, presents a comprehensive overview of machine learning (ML) methods for analysing genetically complex diseases, including MS. I briefly described the mathematical functions that transform classical logistic regression model into ML models for genetic risk prediction. Using case-control genotype data from the Ausimmune Study, I then compared the predictive accuracies of some selected variable selection methods in predicting MS risk. I found that ML algorithms particularly Least Absolute Shrinkage and a Selection Operator (LASSO), Ridge Regression (RIDGE), Elastic Net (ENET), Smoothly Clipped Absolute Deviation (SCAD) penalty, Random Forest (RF) and others, are already being applied to address a variety of important questions in complex genetic diseases, including MS. This study highlights the importance of ML algorithms in the prediction of genetically complex diseases.
In the second study, I examined associations between known clinical, environmental, and genetic factors and the risk of developing new relapses and disability worsening events in people living with MS. By combining the effects of these factors, I developed a time-dynamic clinical-environmental-genotypic prognostic index (CEGPI) and predicted the probability of developing new relapses and disability worsening outcomes. The results from this study suggest that genetic variants have a time-dynamic effects on MS outcomes. Five-year dynamic risk profiles using the developed index revealed good prognostic and discriminative capabilities in terms of risk stratification.
In the third study, I used the time-dynamic genetic prognostic indices developed from the second study and examined the indirect contributions of relapses and treatment to the risk of worsening of disability, and vice-versa. I found that early relapses within 2.5 years of FDE predicted early disability worsening outcomes but did not contribute to long-term disability accrual thereinafter. Conversely, disability worsening events contributed to relapse risk significantly each year and persisted over time regardless of treatment effects. The use of DMTs significantly reduced the hazards associated with these endpoints. This study provided evidence that in early relapse-onset MS, worsening of disability 2.5 years post FDE occurs in ways not clearly tied to relapses, and is strongly linked to an increased risk of future relapses. However, early in the disease course, there is strong evidence for a link between relapses and disability worsening events which can be significantly mitigated by early treatment with DMTs.
he fourth study investigated independent associations between 208 previously established MS genetic loci and the risk worsening of disability over time. Specifically, I developed and validated a robust ensemble ML method to identify people with MS at risk of future worsening. After functional annotation and gene enrichment analysis, I found 7 independent genetic loci intronic or intergenic to genes implicated in peptide hormones and steroids biosynthesis pathways. These variants explained respectively, ∼49%, ∼43%, ∼36%, and ∼42% of the differences in disability progression rates between individuals with clinically isolated syndrome (CIS), relapse?onset MS (ROMS), secondary progressive MS (SPMS), and progressive onset MS (POMS). The derived ensemble models produced a set of genetic decision rules that can be translated to provide additional prognostic values to existing clinical predictions, with the additional benefit of incorporating relevant genetic information into clinical decision making for people with MS. This study extended our knowledge of MS progression genetics and provided evidence that MS related genetic variants have significant effects on MS outcomes.
In the fifth study, I conducted a multi-state genome-wide transition analysis to identify genetic variants having significant effects on the instantaneous risk to transit between disability milestones (defined on the expanded disability states scale scores). I found 36 loci having significant and meaningful influence on the transition hazards and transition probabilities between disability states. These loci were relatively common (MAF>10%) and explained ∼47% of the total liability in disability progression that is due to heritable genetic factors. To the best of my knowledge, this represents the highest explained variance in disability progression attributable to common variants not associated with MS risk. The clinical and environmental predictors examined particularly sex, age at clinical visit, latitudes, and MS course at visit, explained only ∼20% of the phenotypic variance. Together, both components explained ∼75% of the total variability, with the genetic variants being better prognostic factors overall (P< 1x10-16). Further, the integration analysis revealed 10 independent lead variants intronic or intergenic to genes expressed in EBV-transformed lymphocytes and implicating CNS processes including neurotransmission, regulation of (innate) immune response and development, and regulation of synapse activity, and structural integrity of neurons. This study presents overwhelming evidence that MS disability progression is under strong genetic control.
In sum, we examined genetic and environmental determinants involved in the development and progression of disability in MS. Using novel statistical methodologies, we develop robust and clinically impactful predictions regarding an individual’s MS disease time course. In addition to the major roles that clinical and environmental factors both play in MS onset and disability progression, genetic variations, particularly single nucleotide polymorphisms, have additional prognostic capabilities in predicting disease time course. Depending on the outcomes analysed, we observed that MS disease progression is driven by different genetic variants. Specifically, we observed only 8% overall between MS-related genetic variations affecting MS relapses and those associated with disability worsening outcomes. We also observed that the effects of genetic variants on disease progression may change over time depending on the environmental exposures. Including these time dynamic components of genetic variations in prognostic models improved their sensitivities in predicting disease time course. In conclusion, disease progression in MS is strongly influenced by genetic variants and clinical and environmental factors.